Seismic and well log data integration using data-matching techniques

Abstract

Relating well log data to seismic data is an important step in integrated reservoir characterization studies. Traditionally, an interpreter uses well log data, which has high vertical resolution but little lateral coverage, to understand amplitude variations in seismic data, which has lower vertical resolution than well logs but high spatial coverage. The process of calibration is referred to as a seismic-well tie. Several problems arise with the assumptions of conventional seismic-well tie workflows. The seismic-well tie involves generating a reflectivity series from available sonic and density logs acquired at the well, which inherently assumes all wells have a sonic and density log available along the entire length of the well. In many cases, this assumption is not valid as the number of wells drilled often out-numbers the number of sonic and density logs acquired. Common procedures to account for missing well logs in seismic-well ties are to use a time-depth relationship from a nearby well or use an empirical relationship to estimate the missing well log from an available well log. These methods provide constructive solutions. However, variations in structure, stratigraphy or missing/incomplete well logs can result in inaccurate seismic-well ties. In this thesis, I propose a method that predicts missing well log data by first estimating the shifts that align well logs with a reference type log. Once in this stratigraphically correlated, or `relative geologic time,' domain, I interpolate the missing well log data from available logs of the same type. The resulting well log is consistent with available well data and is not distorted by structural or stratigraphic variations. Once complete well log suites are estimated for each well, I focus on improving the efficiency and consistency of multiple seismic-well ties. The seismic-well tie typically involves a subjective and labor-intensive workflow that depends on the interpreter's experience and intuition. I introduce an automatic workflow using local similarity to match the synthetic with the real seismic trace. The advantage of using local similarity to compute the seismic-well tie is that consistent, repeatable, seismic-well ties are achieved. I generate a global log property volume by interpolating log data along local seismic structure and perform blind well tests to verify the accuracy and consistency of seismic-well ties. I apply this workflow to a 3D seismic dataset with 26 wells and achieve consistent, accurate and reproducible seismic-well ties. Combining the results of the well log interpolation and seismic-well tie I can generate a time-to-depth relationship for each well regardless of the initial well log suite. As a result, it is possible to generate log property volumes that integrate the high spatial coverage of seismic data with information from well log data. Well log data can also provide a useful source of information during velocity model building for depth migration. Using concepts and workflows described previously, I show that the mis-tie between a modeled synthetic and real seismic trace is related to an inaccurate migration velocity. Furthermore, this information can be used to update the migration velocity model such that modeled synthetic seismograms, the seismic image, migration velocities and well log velocities become consistent.